Garrett, M.; Carnwath, T.; Albano, R.; Zhuang, Y.; Behrmann, C. A.; Pemberton, M.; Barakat, F.; Lober, R.; Hoeprich, M.; Paravati, A.; Reed, M.; Spry, H.; Woo, D.; O'Brien, E.; VanCauwenbergh, B.; Perentesis, J.; Nasser, R.; Medvedovic, M.; Plas, D.

Introduction: Glioblastomas utilize malignant gene expression pathways to drive growth. Many of these gene pathways are not directly accessible with molecularly targeted pharmacological agents. Chromatin-modifying compounds can alter gene expression and target glioblastoma growth pathways. In this study, we utilize a systematic screen of chromatin-modifying compounds on a panel of patient-derived glioblastoma lines to identify promising compounds and their associated gene targets. Methods: Five glioblastoma cell lines were subjected to a drug screen of 106 chromatin-modifying compounds representing 36 unique drug classes to determine the twelve most promising drug classes and the best candidate inhibitors in each class. These twelve drugs were then tested with a panel of twelve patient-derived gliomasphere lines to identify growth inhibition and corresponding gene expression patterns. Overlap analysis and weighted co-expression network analysis (WCGNA) were utilized to determine potential target genes and gene pathways. Results: The initial drug screen identified twelve candidate pharmacologic agents for further testing. Drug sensitivity testing indicated an overall high degree of variability between gliomasphere lines. However, CPI203 was the most consistently effective compound, and the BET inhibitor class was the most consistently effective class of compounds across the gliomasphere panel. Correspondingly, most of the compounds tested had highly variable effects on gene expression between gliomasphere lines. CPI203 stood out as the only compound to induce a consistent effect on gene expression across different gliomasphere lines, specifically down-regulation of DNA-synthesis genes. Amongst the twelve tested cell lines, high expression of CDKN2A and CDKN2B distinguished more drug sensitive from more drug resistant lines. WCGNA identified two oncogenic gene modules (FBXO5 and MELK) that were effectively downregulated by CPI203 (FBXO5) and ML228 (FBXO5 and MELK). Conclusions: The bromodomain inhibitor CPI203 induced relatively consistent effects on gene expression and growth across a variety of glioblastoma lines, specifically down-regulating genes associated with DNA replication. We propose that clinically effective BET inhibitors have the potential to induce consistent beneficial effects across a spectrum of glioblastomas.